* the configurational theories
assuming a strong influence of the basic order plus semantic/
pragmatic reasons for reordering (e.g., GB, minimalism)
* the Prague School assuming information structure as the ordering
principle except for the focus constituents where the systemic
order prevails (Prague School)
* a performance-based theory based on length (Hawkinss EIC-"Early
Immediate Constituent Principle")
Recently Hawkinss (1994) length-based theory has received much
attention in general linguistics (typology), computational
linguistics (modelling language evolution) and psycholinguistics
(memory-based models of sentence processing). According to Hawkins,
the influence of the other factors besides length can almost
entirely be explained as epiphenomena of length. The claim is that
all types of phrases that tend to precede their siblings, such as
topic phrases, pronominal NPs, complements preceding in basic order
and definite NPs are simply shorter on the average than their
respective counterparts. The data presented by Hawkins himself are
suggestive but much too restricted in size to permit any
empirically supported conclusions.
In this paper we report on a corpus-based study involving six
German verbs exhibiting different basic-order patterns (SUBJSUBJ,
DOBJDOBJ). In order to prohibit the effects of other factors, we
restricted the investigation to non-pronominal NPs in the middle
field. Because of the low frequency of any individual verb, a
corpus of 32 Mio words (ca. 1.6 Mio sentences) had to be chosen.
The corpus consists of contemporary newspaper texts. For three
verbs each, pairs of SUBJ-IOBJ and DOBJ-IOBJ were considered. The
aim was to search for any interdependence of basic order and length
and in this context to also investigate the role of definiteness.
The search was to a large part automated.
Results: Hawkins's claim that the basic order is a mere
epiphenomenon of length cannot be maintained. EIC is not a primary
factor for the order among verbal complements. Also: contrary to
Hawkinss prediction, there is no correlation between verbal
position (before or after the middle field) and EIC. But there is a
strong independent influence of basic order. There is also an
interesting effect concerning definiteness: In addition to the
expected preference definiteP>